Summary

  • CoreWeave reported $2,575 million of revenue in the second quarter of 2026 against roughly $104 billion of revenue backlog — about 2.5% of the closing backlog recognised in the quarter, down from about 4.0% a year earlier.
  • Net interest expense rose to $640 million from $267 million year over year, and the company raised its 2026 capital expenditure plan to $35–39 billion, so the delivery chain is now financed at a cost that grows faster than recognised revenue.
  • The company's own definition of backlog is conditional on delivery and service availability; nothing in the disclosures shows contracted demand falling short, and nothing shows the energization, capital and financing steps keeping pace.

What CoreWeave actually reported

For the three months ended 30 June 2026, CoreWeave reported revenue of $2,575 million, up from $1,212 million in the same quarter of 2025 (company release; SEC exhibit). The same release shows an operating loss of $(49) million against operating income of $19 million a year earlier, and a net loss of $(626) million versus $(290) million, a net loss margin of (24)%. Adjusted EBITDA was $1,510 million at a 59% margin, down from a 62% margin, and adjusted operating income was $128 million, a 5% margin.

Revenue backlog was approximately $104 billion as of 30 June 2026 (company release). The company states that this figure excludes more than $25 billion of net new customer commitments added in early Q3, and that backlog "includes remaining performance obligations, plus other amounts we estimate will be recognized as revenue in future periods under committed customer contracts, in each case, subject to the satisfaction of delivery and availability of service requirements." Its earnings presentation puts the composition at $103.7 billion of remaining performance obligations plus $0.5 billion of other estimated future revenue, up 246% year over year (2Q26 presentation).

The useful measurement is the ratio between the two: quarterly revenue divided by closing backlog was about 4.0% in Q2 2025 ($1,212.8 million against $30.1 billion) and about 2.5% in Q2 2026 ($2,575 million against roughly $104 billion) (Q2 2025 release). A backlog that grows faster than the revenue recognised against it is not automatically a demand problem; it is a schedule problem, and the schedule is set by power, buildings and financing.

The three steps that bind

The sequence is straightforward. A contract is signed. Data-centre capacity is leased or built. Power is energized and the site is interconnected. GPUs are installed and billed. Between mid-2025 and mid-2026 the backlog series moved $30.1 billion (30 June 2025), $55.6 billion (30 September 2025), $66.8 billion (31 December 2025), $99.4 billion (31 March 2026, per Reuters) and about $104.2 billion (30 June 2026), while quarterly revenue moved $1,212.8 million, $1,364.7 million, $1,572 million and then $2,575 million (Q2 2025; Q3 2025; Q4 and FY2025; Reuters).

The financing step is the one with the clearest published cost. Net interest expense was $(640) million in Q2 2026 against $(267) million a year earlier — an increase of roughly 140%, and equal to about 24.9% of quarterly revenue versus about 22.0% a year before (company release). Capital expenditure was $9,352 million in the quarter and $16,139 million in the first half, per the company's presentation (2Q26 presentation); that quarterly spend was roughly 3.6 times revenue, against about 2.4 times a year earlier, when capital expenditure was $2.9 billion against $1,212.8 million of revenue (Q2'25 presentation).

The energization step is reported at a coarser grain. CNBC reported the company citing 1.5 gigawatts of active power, Q3 2026 revenue guidance of $3.4–3.6 billion, full-year 2026 revenue guidance of $12.4–13.2 billion, and a target of more than 1.85 gigawatts of active power by year-end (CNBC). Reuters reported the capital expenditure plan raised to $35–39 billion from $31–35 billion (Reuters). Those two disclosures point the same way: the company is spending more, sooner, to reach a power figure that is still under two gigawatts against a contracted book far larger than one year of revenue.

Evidence boundaries

The figures above come from company releases, an SEC exhibit, an investor deck and two news reports. The underlying filings were not read directly here, and tabular values were retrieved as reported excerpts; the Q4 and fiscal-year 2025 table in particular was returned in fragmented form. Revenue backlog is a company-defined measure, and the retrieved material does not disclose its duration profile, its customer mix, or the split between capacity already under construction and capacity not yet permitted.

No independent utility, interconnection-queue or data-centre delivery record was obtained, so the energization constraint is described from company-reported active-power figures and guidance rather than third-party delivery data. Guidance figures are forecasts and can be revised. Nothing in the retrieved material shows a customer terminating, renegotiating or writing down a commitment.

Working reading

On the evidence available, contracted demand is not the binding constraint at CoreWeave. The steps that are visible, quantified and growing fastest are the cost of capital and the pace of capital deployment needed to energize capacity; the conversion of backlog into revenue is a function of those, and the company's own definition of backlog concedes the conditionality. The honest conclusion is bounded: if energization and financing stay on the disclosed trajectory, backlog conversion stays slow by arithmetic even while absolute revenue compounds. If they do not, the constraint moves somewhere this evidence set cannot see.

Sources